An agent for work behind the sale
On 12 November, Square made Squaremates generally available to US sellers. The agent builds on the earlier Managerbot and, according to the company, becomes the entry point to the Square Dashboard. Sellers can ask questions, see what needs attention and delegate specific work. Square notes an exception to general availability for users of Multi-Account Suite.
Examples range from analysing past sales and drafting a campaign for returning customers to changing catalogue prices, descriptions and images. Square also says the agent can compare labour costs with sales and prepare a staffing schedule for review. These are the provider’s descriptions; their reliability in an individual business needs to be assessed in use.
From low stock to a prepared order
A particularly clear example is a restocking feature that Square still labels beta. The agent can identify low stock, prepare an order and build it on a supplier’s website. The seller completes the order. The task therefore crosses into another service, although the announcement does not establish autonomous purchasing or payment.
A café might ask it to check which cups are running low. Before approving an order, the owner still needs answers: is the recorded stock right, what minimum quantity does the supplier require, and do the price and delivery date fit? A useful proposal makes those grounds visible so the owner does not have to repeat the entire search.
Connecting other tools to Square
Alongside the agent, Square announced an improved MCP interface and a new command line interface. Through MCP, external assistants can work with a connected Square account and a defined set of actions. The command line tool is aimed at sellers and developers who want to script or schedule work, or build applications using Square data. According to Square, Squaremates Apps can be deployed and hosted on its platform.
This gives businesses more choice in how they work. It also makes permissions more concrete: who can read data, change a price or trigger a prepared action? A connection alone does not determine which limits are appropriate for a particular business.
What this means for store operations
In chapter 11 of my book, I describe a progression from a specific task with approval towards an operator that handles recurring work within defined limits. Squaremates shows how many practical tasks can fit into that starting point. The announced examples do not by themselves establish persistent, objective-led operation.
Replit’s agent for purchasing and advertising also brings operational work and merchant decisions together. Square starts with catalogue, stock and staffing. In both cases the useful questions are which information the agent actually uses and where an action becomes binding.
A useful first test
A team can start with a recurring task whose result is easy to check. For restocking, that might mean a small product group, a familiar supplier and a comparison against actual stock. The team can then judge whether the proposal saves time and whether price, quantity and approval fit reliably into its own process.
Source and further reading
Read the original announcement ↗